A Fibonacci retracement measures how much of a selected price move has been given back. Choose two anchors and an explicit calculation convention first. The grid alone supplies no bounce probability, next-candle direction or appropriate order price; its precision depends on those inputs and settings.

What a Fibonacci retracement actually is

For the stated linear model, let A begin the move, B end it and r denote the retraced fraction. Define P(r) = B + r(A − B): zero retracement is at B and 100% at A. After a rise from L to H, P = H − r(H − L); after a decline H to L, P = L + r(H − L). This measures a share of that range, not a percentage loss in the whole instrument’s price.

Retracement and projection perform different calculations, but their names do not prescribe exclusively entries or exits. MT5 and TradingView allow levels outside the two-anchor range. The separate Fibonacci levels for setting targets lesson compares reference points and two- versus three-anchor constructions. Record the zero/100% prices rather than infer the convention from click order alone.

Where the 38.2%, 50% and 61.8% levels come from

In the sequence 0, 1, 1, 2, 3, 5, 8, 13… each later term is the sum of the preceding two. As terms become larger, dividing one by the next, by the term two places ahead or by the term three places ahead approaches about 0.618, 0.382 and 0.236. Initial ratios need not already equal these limits. The 78.6% convention rounds the square root of about 0.618; 1.618 approximates the golden ratio and 0.618 its inverse.

Half the range, 50%, is an additional convention rather than one of those limiting Fibonacci ratios. Its inclusion does not give it a measured psychological strength or a guaranteed market reaction. Mathematical origins of the other coefficients likewise do not establish their trading efficacy.

Fractions of a stated range, not success probabilities
23.6%r = 0.236 of the selected move.
38.2%r = 0.382; the input anchors still determine the price.
50%Half the range, an added convention.
61.8%r = 0.618, not a 61.8% chance of reversal.
78.6%A rounded coefficient; crossing it does not itself establish a new trend.

Average first, or invert the quote first?

Hypothetical EUR/USD rises from A=1.10000 to B=1.20000 USD per EUR. Inverting units gives a fall from A′=1/A to B′=1/B EUR per USD. On both linear axes, calculate a 50% retracement from the movement's end.

Scroll sideways to see the full diagram

Average first, or invert the quote first?The midpoint of 1.10000 to 1.20000 is 1.15000 USD per EUR; its inverse is about 0.869565 EUR per USD. Inverted anchors are about 0.909091 and 0.833333 EUR per USD, whose midpoint is about 0.871212. The average of inverses differs from the inverse of the average. Labels are rounded only after calculation.Inverting a pair changes its linear midpointUSD per 1 EUREUR per 1 USD50%50%B: 1.200001.15000A: 1.10000A′: 0.9090910.871212B′: 0.833333Inverse of left midpoint: 1 / 1.15000= 0.869565 EUR per USDMean of inverses: (1/1.1 + 1/1.2) / 2= 0.871212 EUR per USD
Assumptions: positive reference prices, no Bid/Ask or spread; not real quotes or order prices. Both directions use P(r)=B+r(A−B), r=0.5; inverted anchors A′=1/A and B′=1/B. Before rounding, 1/1.15≈0.8695652174 and (1/1.1+1/1.2)/2≈0.8712121212 EUR per USD. Units and geometry change without new information about future prices. CME: base and quote currency. TradingView: anchors and grid direction.
A linear retracement does not preserve the same price after quote inversion.

How to draw the tool step by step

TradingView documents precise coordinates, custom ratios and Reverse; it also offers a log-based calculation when the chart uses logarithmic scale. Identical anchors without identical settings need not give identical lines. To reproduce this lesson, use linear calculation with 0% at B, and record instrument, feed, interval and wick or close anchors.

Anchor selection must follow a stated rule, without a universal requirement for hundreds of pips or a daily chart. Changing the selected high from 1.0950 to 1.1000 with low 1.0750 fixed makes the 61.8% line 1.1000 − 0.618 × 0.0250 = 1.08455, which is 19.1 pips above the earlier 1.08264. Choosing a new point after observing a later reaction revises the scenario rather than validate the original one.

A reference level is separate from a trading decision

A grid does not determine size, fill, cost or exit. Confluence with prior support or resistance can define an additional condition to test, but another average or candle pattern may process the same prices. Coincidence alone is not independent evidence or a guaranteed improvement in net performance.

A 61.8% touch is not an automatic entry, nor does crossing 78.6% establish that a trend must reverse. State what each threshold means in the selected plan before collecting outcomes. Ordinary stops may execute worse than their stated level. Trend-following strategies discuss separate rule choices; their names do not supply an optimal buffer or a loss cap.

Hypothetical 200-pip range in both directions

Assume hypothetical EUR/USD with pip = 0.0001 USD per EUR, L = 1.0750 and H = 1.0950. The difference 0.0200 is 200 pips. These are two separate scenarios: a correction after L→H and a bounce after H→L, without a future outcome or recommended trade.

Linear retracements of the same hypothetical range
rRange portion, pipsAfter L→HAfter H→L
23.6%47.21.090281.07972
38.2%76.41.087361.08264
50%1001.085001.08500
61.8%123.61.082641.08736
78.6%157.21.079281.09072

For 61.8% after the rise, 1.0950 − 0.618 × 0.0200 = 1.08264; after the decline, 1.0750 + 0.618 × 0.0200 = 1.08736. Only the 50% midpoint is shared. Giving back half of this 200-pip move does not mean the euro lost 50% of its value. Keep calculation precision before fitting any order to the instrument’s price step.

Retracement levels are calculated from selected anchors; a ratio alone is not an entry signal.

The limits of the method and the most common mistakes

Fibonacci levels depend on selected price anchors and do not compel an exchange rate to reverse. Claims that orders cluster at these levels or cause reactions require relevant market evidence; popularity alone does not establish effectiveness.

Two nearly matching grids may even repeat the same algebra. In an own hypothetical counterexample, a large move is A = 100 to B = 120. Deliberately set the smaller move’s end to 100 + 0.618 × 20 = 112.36, with the same start. Large-move 61.8% is 107.64; smaller-move 38.2% is 112.36 − 0.382 × 12.36 = 107.63848, only 0.00152 away. These are not two independently selected market extremes.

The rounded 0.618² = 0.381924 almost equals 0.382. For exact q = (√5 − 1)/2, q² = 1 − q, so suitably nested ranges reproduce the same line exactly. This does not show that every confluence is artificial; it shows why anchor origin and independent selection matter. Close lines alone add no measured bounce probability.

What to do tomorrow

  1. Calculate both directions. Reproduce the 200-pip table with 0% at the selected endpoint B and compare 61.8% prices.
  2. Check settings. Record exact anchors, source, interval, Reverse and linear versus logarithmic calculation before comparing outputs.
  3. Change one anchor. Keep L = 1.0750 and set H = 1.1000. Upward 38.2% becomes 1.09045 and 61.8% 1.08455; their separation is 0.00590, or 59 pips.
  4. Preserve availability. If a peak needs two later closed bars for confirmation, do not use that final peak on its own bar in a test.
  5. Evaluate the complete rule. Keep cases without a bounce, execution costs and unused data. A month of chosen examples is not automatically sufficient evidence of superiority.

Related reading: Fibonacci extensions for setting targets compares projection origins; drawing support and resistance helps specify a context condition; trend-following systems addresses separately defined rules, and harmonic patterns develops ratio-based constructions. For broader context, see the technical analysis section on ForexMechanics.com.